refactor: Simplify Gemini function parameter handling

- Removed redundant checks for non-empty properties in function parameters.
- Set function parameters to nil when no properties are needed, streamlining the logic for handling Gemini requests.
- Improved code clarity and maintainability by eliminating unnecessary complexity.

README.en.md

@@ -81,6 +81,7 @@ You can add custom models gpt-4-gizmo-* in channels. These are third-party model
 - `UPDATE_TASK`: Update async tasks (Midjourney, Suno), default `true`
 - `GEMINI_MODEL_MAP`: Specify Gemini model versions (v1/v1beta), format: "model:version", comma-separated
 - `COHERE_SAFETY_SETTING`: Cohere model [safety settings](https://docs.cohere.com/docs/safety-modes#overview), options: `NONE`, `CONTEXTUAL`, `STRICT`, default `NONE`
+- `GEMINI_VISION_MAX_IMAGE_NUM`: Gemini model maximum image number, default `16`, set to `-1` to disable
 
 ## Deployment
 > [!TIP]

README.md

@@ -87,6 +87,7 @@
 - `UPDATE_TASK`：是否更新异步任务（Midjourney、Suno），默认为 `true`，关闭后将不会更新任务进度。
 - `GEMINI_MODEL_MAP`：Gemini模型指定版本(v1/v1beta)，使用“模型:版本”指定，","分隔，例如：-e GEMINI_MODEL_MAP="gemini-1.5-pro-latest:v1beta,gemini-1.5-pro-001:v1beta"，为空则使用默认配置(v1beta)
 - `COHERE_SAFETY_SETTING`：Cohere模型[安全设置](https://docs.cohere.com/docs/safety-modes#overview)，可选值为 `NONE`, `CONTEXTUAL`，`STRICT`，默认为 `NONE`。
+- `GEMINI_VISION_MAX_IMAGE_NUM`：Gemini模型最大图片数量，默认为 `16`，设置为 `-1` 则不限制。
 ## 部署
 > [!TIP]
 > 最新版Docker镜像：`calciumion/new-api:latest`  

common/model-ratio.go

@@ -46,6 +46,8 @@ var defaultModelRatio = map[string]float64{
 	"gpt-4o-2024-08-06":               1.25, // $2.5 / 1M tokens
 	"gpt-4o-2024-11-20":               1.25, // $2.5 / 1M tokens
 	"gpt-4o-realtime-preview":         2.5,
+	"o1":                              7.5,
+	"o1-2024-12-17":                   7.5,
 	"o1-preview":                      7.5,
 	"o1-preview-2024-09-12":           7.5,
 	"o1-mini":                         1.5,

constant/env.go

@@ -23,6 +23,8 @@ var GeminiModelMap = map[string]string{
 	"gemini-1.0-pro": "v1",
 }
 
+var GeminiVisionMaxImageNum = common.GetEnvOrDefault("GEMINI_VISION_MAX_IMAGE_NUM", 16)
+
 func InitEnv() {
 	modelVersionMapStr := strings.TrimSpace(os.Getenv("GEMINI_MODEL_MAP"))
 	if modelVersionMapStr == "" {

dto/openai_request.go

@@ -3,39 +3,48 @@ package dto
 import "encoding/json"
 
 type ResponseFormat struct {
-	Type string `json:"type,omitempty"`
+	Type       string            `json:"type,omitempty"`
+	JsonSchema *FormatJsonSchema `json:"json_schema,omitempty"`
+}
+
+type FormatJsonSchema struct {
+	Description string `json:"description,omitempty"`
+	Name        string `json:"name"`
+	Schema      any    `json:"schema,omitempty"`
+	Strict      any    `json:"strict,omitempty"`
 }
 
 type GeneralOpenAIRequest struct {
-	Model               string         `json:"model,omitempty"`
-	Messages            []Message      `json:"messages,omitempty"`
-	Prompt              any            `json:"prompt,omitempty"`
-	Stream              bool           `json:"stream,omitempty"`
-	StreamOptions       *StreamOptions `json:"stream_options,omitempty"`
-	MaxTokens           uint           `json:"max_tokens,omitempty"`
-	MaxCompletionTokens uint           `json:"max_completion_tokens,omitempty"`
-	Temperature         float64        `json:"temperature,omitempty"`
-	TopP                float64        `json:"top_p,omitempty"`
-	TopK                int            `json:"top_k,omitempty"`
-	Stop                any            `json:"stop,omitempty"`
-	N                   int            `json:"n,omitempty"`
-	Input               any            `json:"input,omitempty"`
-	Instruction         string         `json:"instruction,omitempty"`
-	Size                string         `json:"size,omitempty"`
-	Functions           any            `json:"functions,omitempty"`
-	FrequencyPenalty    float64        `json:"frequency_penalty,omitempty"`
-	PresencePenalty     float64        `json:"presence_penalty,omitempty"`
-	ResponseFormat      any            `json:"response_format,omitempty"`
-	EncodingFormat      any            `json:"encoding_format,omitempty"`
-	Seed                float64        `json:"seed,omitempty"`
-	Tools               []ToolCall     `json:"tools,omitempty"`
-	ToolChoice          any            `json:"tool_choice,omitempty"`
-	User                string         `json:"user,omitempty"`
-	LogProbs            bool           `json:"logprobs,omitempty"`
-	TopLogProbs         int            `json:"top_logprobs,omitempty"`
-	Dimensions          int            `json:"dimensions,omitempty"`
-	Modalities          any            `json:"modalities,omitempty"`
-	Audio               any            `json:"audio,omitempty"`
+	Model               string          `json:"model,omitempty"`
+	Messages            []Message       `json:"messages,omitempty"`
+	Prompt              any             `json:"prompt,omitempty"`
+	Stream              bool            `json:"stream,omitempty"`
+	StreamOptions       *StreamOptions  `json:"stream_options,omitempty"`
+	MaxTokens           uint            `json:"max_tokens,omitempty"`
+	MaxCompletionTokens uint            `json:"max_completion_tokens,omitempty"`
+	ReasoningEffort     string         `json:"reasoning_effort,omitempty"`
+	Temperature         float64         `json:"temperature,omitempty"`
+	TopP                float64         `json:"top_p,omitempty"`
+	TopK                int             `json:"top_k,omitempty"`
+	Stop                any             `json:"stop,omitempty"`
+	N                   int             `json:"n,omitempty"`
+	Input               any             `json:"input,omitempty"`
+	Instruction         string          `json:"instruction,omitempty"`
+	Size                string          `json:"size,omitempty"`
+	Functions           any             `json:"functions,omitempty"`
+	FrequencyPenalty    float64         `json:"frequency_penalty,omitempty"`
+	PresencePenalty     float64         `json:"presence_penalty,omitempty"`
+	ResponseFormat      *ResponseFormat `json:"response_format,omitempty"`
+	EncodingFormat      any             `json:"encoding_format,omitempty"`
+	Seed                float64         `json:"seed,omitempty"`
+	Tools               []ToolCall      `json:"tools,omitempty"`
+	ToolChoice          any             `json:"tool_choice,omitempty"`
+	User                string          `json:"user,omitempty"`
+	LogProbs            bool            `json:"logprobs,omitempty"`
+	TopLogProbs         int             `json:"top_logprobs,omitempty"`
+	Dimensions          int             `json:"dimensions,omitempty"`
+	Modalities          any             `json:"modalities,omitempty"`
+	Audio               any             `json:"audio,omitempty"`
 }
 
 type OpenAITools struct {

relay/channel/gemini/adaptor.go

@@ -57,7 +57,11 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, re
 	if request == nil {
 		return nil, errors.New("request is nil")
 	}
-	return CovertGemini2OpenAI(*request), nil
+	ai, err := CovertGemini2OpenAI(*request)
+	if err != nil {
+		return nil, err
+	}
+	return ai, nil
 }
 
 func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {

relay/channel/gemini/constant.go

@@ -1,14 +1,18 @@
 package gemini
 
-const (
-	GeminiVisionMaxImageNum = 16
-)
-
 var ModelList = []string{
+	// stable version
+	"gemini-1.5-pro", "gemini-1.5-flash", "gemini-1.5-flash-8b",
+	// latest version
 	"gemini-1.5-pro-latest", "gemini-1.5-flash-latest",
+	// legacy version
 	"gemini-1.5-pro-exp-0827", "gemini-1.5-flash-exp-0827",
+	// exp
 	"gemini-exp-1114", "gemini-exp-1121", "gemini-exp-1206",
+	// flash exp
 	"gemini-2.0-flash-exp",
+	// thinking exp
+	"gemini-2.0-flash-thinking-exp",
 	"gemini-2.0-flash-thinking-exp-1219",
 }
 

relay/channel/gemini/dto.go

@@ -40,12 +40,14 @@ type GeminiChatTools struct {
 }
 
 type GeminiChatGenerationConfig struct {
-	Temperature     float64  `json:"temperature,omitempty"`
-	TopP            float64  `json:"topP,omitempty"`
-	TopK            float64  `json:"topK,omitempty"`
-	MaxOutputTokens uint     `json:"maxOutputTokens,omitempty"`
-	CandidateCount  int      `json:"candidateCount,omitempty"`
-	StopSequences   []string `json:"stopSequences,omitempty"`
+	Temperature      float64  `json:"temperature,omitempty"`
+	TopP             float64  `json:"topP,omitempty"`
+	TopK             float64  `json:"topK,omitempty"`
+	MaxOutputTokens  uint     `json:"maxOutputTokens,omitempty"`
+	CandidateCount   int      `json:"candidateCount,omitempty"`
+	StopSequences    []string `json:"stopSequences,omitempty"`
+	ResponseMimeType string   `json:"responseMimeType,omitempty"`
+	ResponseSchema   any      `json:"responseSchema,omitempty"`
 }
 
 type GeminiChatCandidate struct {

relay/channel/gemini/relay-gemini.go

@@ -17,7 +17,7 @@ import (
 )
 
 // Setting safety to the lowest possible values since Gemini is already powerless enough
-func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) *GeminiChatRequest {
+func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) (*GeminiChatRequest, error) {
 	geminiRequest := GeminiChatRequest{
 		Contents: make([]GeminiChatContent, 0, len(textRequest.Messages)),
 		SafetySettings: []GeminiChatSafetySettings{
@@ -56,6 +56,16 @@ func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) *GeminiChatReques
 				googleSearch = true
 				continue
 			}
+			if tool.Function.Parameters != nil {
+				params, ok := tool.Function.Parameters.(map[string]interface{})
+				if ok {
+					if props, hasProps := params["properties"].(map[string]interface{}); hasProps {
+						if len(props) == 0 {
+							tool.Function.Parameters = nil
+						}
+					}
+				}
+			}
 			functions = append(functions, tool.Function)
 		}
 		if len(functions) > 0 {
@@ -77,6 +87,15 @@ func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) *GeminiChatReques
 			},
 		}
 	}
+	if textRequest.ResponseFormat != nil && (textRequest.ResponseFormat.Type == "json_schema" || textRequest.ResponseFormat.Type == "json_object") {
+		geminiRequest.GenerationConfig.ResponseMimeType = "application/json"
+
+		if textRequest.ResponseFormat.JsonSchema != nil && textRequest.ResponseFormat.JsonSchema.Schema != nil {
+			cleanedSchema := removeAdditionalPropertiesWithDepth(textRequest.ResponseFormat.JsonSchema.Schema, 0)
+			geminiRequest.GenerationConfig.ResponseSchema = cleanedSchema
+		}
+	}
+
 	//shouldAddDummyModelMessage := false
 	for _, message := range textRequest.Messages {
 
@@ -108,9 +127,10 @@ func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) *GeminiChatReques
 				})
 			} else if part.Type == dto.ContentTypeImageURL {
 				imageNum += 1
-				//if imageNum > GeminiVisionMaxImageNum {
-				//	continue
-				//}
+
+				if constant.GeminiVisionMaxImageNum != -1 && imageNum > constant.GeminiVisionMaxImageNum {
+					return nil, fmt.Errorf("too many images in the message, max allowed is %d", constant.GeminiVisionMaxImageNum)
+				}
 				// 判断是否是url
 				if strings.HasPrefix(part.ImageUrl.(dto.MessageImageUrl).Url, "http") {
 					// 是url，获取图片的类型和base64编码的数据
@@ -124,7 +144,7 @@ func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) *GeminiChatReques
 				} else {
 					_, format, base64String, err := service.DecodeBase64ImageData(part.ImageUrl.(dto.MessageImageUrl).Url)
 					if err != nil {
-						continue
+						return nil, fmt.Errorf("decode base64 image data failed: %s", err.Error())
 					}
 					parts = append(parts, GeminiPart{
 						InlineData: &GeminiInlineData{
@@ -161,7 +181,47 @@ func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) *GeminiChatReques
 		//	shouldAddDummyModelMessage = false
 		//}
 	}
-	return &geminiRequest
+	return &geminiRequest, nil
+}
+
+func removeAdditionalPropertiesWithDepth(schema interface{}, depth int) interface{} {
+	if depth >= 5 {
+		return schema
+	}
+
+	v, ok := schema.(map[string]interface{})
+	if !ok || len(v) == 0 {
+		return schema
+	}
+
+	// 如果type不为object和array，则直接返回
+	if typeVal, exists := v["type"]; !exists || (typeVal != "object" && typeVal != "array") {
+		return schema
+	}
+
+	switch v["type"] {
+	case "object":
+		delete(v, "additionalProperties")
+		// 处理 properties
+		if properties, ok := v["properties"].(map[string]interface{}); ok {
+			for key, value := range properties {
+				properties[key] = removeAdditionalPropertiesWithDepth(value, depth+1)
+			}
+		}
+		for _, field := range []string{"allOf", "anyOf", "oneOf"} {
+			if nested, ok := v[field].([]interface{}); ok {
+				for i, item := range nested {
+					nested[i] = removeAdditionalPropertiesWithDepth(item, depth+1)
+				}
+			}
+		}
+	case "array":
+		if items, ok := v["items"].(map[string]interface{}); ok {
+			v["items"] = removeAdditionalPropertiesWithDepth(items, depth+1)
+		}
+	}
+
+	return v
 }
 
 func (g *GeminiChatResponse) GetResponseText() string {
@@ -236,13 +296,17 @@ func streamResponseGeminiChat2OpenAI(geminiResponse *GeminiChatResponse) *dto.Ch
 	var choice dto.ChatCompletionsStreamResponseChoice
 	//choice.Delta.SetContentString(geminiResponse.GetResponseText())
 	if len(geminiResponse.Candidates) > 0 && len(geminiResponse.Candidates[0].Content.Parts) > 0 {
-		respFirst := geminiResponse.Candidates[0].Content.Parts[0]
-		if respFirst.FunctionCall != nil {
+		respFirstParts := geminiResponse.Candidates[0].Content.Parts
+		if respFirstParts[0].FunctionCall != nil {
 			// function response
 			choice.Delta.ToolCalls = getToolCalls(&geminiResponse.Candidates[0])
 		} else {
 			// text response
-			choice.Delta.SetContentString(respFirst.Text)
+			var texts []string
+			for _, part := range respFirstParts {
+				texts = append(texts, part.Text)
+			}
+			choice.Delta.SetContentString(strings.Join(texts, "\n"))
 		}
 	}
 	var response dto.ChatCompletionsStreamResponse

relay/channel/openai/adaptor.go

@@ -109,12 +109,19 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, re
 	if info.ChannelType != common.ChannelTypeOpenAI {
 		request.StreamOptions = nil
 	}
-	if strings.HasPrefix(request.Model, "o1-") {
+	if strings.HasPrefix(request.Model, "o1") {
 		if request.MaxCompletionTokens == 0 && request.MaxTokens != 0 {
 			request.MaxCompletionTokens = request.MaxTokens
 			request.MaxTokens = 0
 		}
 	}
+	if request.Model == "o1" || request.Model == "o1-2024-12-17" {
+		//修改第一个Message的内容，将system改为developer
+		if len(request.Messages) > 0 && request.Messages[0].Role == "system" {
+			request.Messages[0].Role = "developer"
+		}
+	}
+
 	return request, nil
 }
 

relay/channel/openai/constant.go

@@ -13,6 +13,7 @@ var ModelList = []string{
 	"gpt-4o-mini", "gpt-4o-mini-2024-07-18",
 	"o1-preview", "o1-preview-2024-09-12",
 	"o1-mini", "o1-mini-2024-09-12",
+	"o1", "o1-2024-12-17",
 	"gpt-4o-audio-preview", "gpt-4o-audio-preview-2024-10-01",
 	"gpt-4o-realtime-preview", "gpt-4o-realtime-preview-2024-10-01",
 	"text-embedding-ada-002", "text-embedding-3-small", "text-embedding-3-large",

relay/channel/vertex/adaptor.go

@@ -135,7 +135,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, re
 		c.Set("request_model", request.Model)
 		return vertexClaudeReq, nil
 	} else if a.RequestMode == RequestModeGemini {
-		geminiRequest := gemini.CovertGemini2OpenAI(*request)
+		geminiRequest, err := gemini.CovertGemini2OpenAI(*request)
+		if err != nil {
+			return nil, err
+		}
 		c.Set("request_model", request.Model)
 		return geminiRequest, nil
 	} else if a.RequestMode == RequestModeLlama {

service/log_info_generate.go

@@ -12,6 +12,7 @@ func GenerateTextOtherInfo(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, m
 	other["group_ratio"] = groupRatio
 	other["completion_ratio"] = completionRatio
 	other["model_price"] = modelPrice
+	other["group"] = relayInfo.Group
 	other["frt"] = float64(relayInfo.FirstResponseTime.UnixMilli() - relayInfo.StartTime.UnixMilli())
 	adminInfo := make(map[string]interface{})
 	adminInfo["use_channel"] = ctx.GetStringSlice("use_channel")

web/src/components/LogsTable.js

@@ -25,7 +25,7 @@ import {
 } from '@douyinfe/semi-ui';
 import { ITEMS_PER_PAGE } from '../constants';
 import {
-  renderAudioModelPrice,
+  renderAudioModelPrice, renderGroup,
   renderModelPrice, renderModelPriceSimple,
   renderNumber,
   renderQuota,
@@ -217,6 +217,29 @@ const LogsTable = () => {
         );
       },
     },
+    {
+      title: t('分组'),
+      dataIndex: 'group',
+      render: (text, record, index) => {
+        if (record.type === 0 || record.type === 2) {
+          let other = JSON.parse(record.other);
+          if (other === null) {
+            return <></>;
+          }
+          if (other.group !== undefined) {
+            return (
+              <>
+                {renderGroup(other.group)}
+              </>
+            );
+          } else {
+            return <></>;
+          }
+        } else {
+          return <></>;
+        }
+      },
+    },
     {
       title: t('类型'),
       dataIndex: 'type',

web/src/pages/Setting/Operation/ModelSettingsVisualEditor.js

@@ -75,7 +75,7 @@ export default function ModelSettingsVisualEditor(props) {
           output.ModelPrice[model.name] = parseFloat(model.price)
         } else {
           if (model.ratio !== '') output.ModelRatio[model.name] = parseFloat(model.ratio);
-          if (model.completionRatio != '') output.CompletionRatio[model.name] = parseFloat(model.completionRatio);
+          if (model.completionRatio !== '') output.CompletionRatio[model.name] = parseFloat(model.completionRatio);
         }
       });
 
@@ -203,11 +203,6 @@ export default function ModelSettingsVisualEditor(props) {
       showError('模型名称已存在');
       return;
     }
-    // 不允许同时添加固定价格和倍率
-    if (values.price !== '' && (values.ratio !== '' || values.completionRatio !== '')) {
-      showError('固定价格和倍率不能同时存在');
-      return;
-    }
     setModels(prev => [{
       name: values.name,
       price: values.price || '',